Earlier quoted context omitted.
CBC mode cannot be parallelized. That is a sequential-only algorithm, and none of the stuff I talked about before can take advantage of CBC mode. GCM mode can be parallelized. Notice AES-128-GCM (which can take advantage of 2x pipelines) really flies on Zen1 compared to 6700.
Regardless, it does really all come down to what your compute need is, and specialized metrics for somewhat odd-ball cases can sometimes still heavily favor Intel for the money. Thanks for your comments! I didn't previously fully comprehend how the AVX modes and pipeline widths mattered for AES.
You don't come up with 32-bytes of random numbers every 4-clock ticks unless you understand this stuff. :-)